WEBVTT NOTE Sentence-level transcript of https://www.youtube.com/watch?v=qJtcEDc_WJU NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/qJtcEDc_WJU.html. A cue ends where the next begins, or 2 s after its last word. s1 00:00:01.309 --> 00:00:03.309 [music] s2 00:00:12.559 --> 00:00:13.040 Okay. s3 00:00:13.040 --> 00:00:15.920 So, yeah, let's let's see what this looks like. s4 00:00:15.920 --> 00:00:20.880 Uh, analyze uh Google and Microsoft. s5 00:00:20.880 --> 00:00:36.559 Um, so ideally what would happen and I hope this works live for y'all um is that uh this main agent is going to um evaluate if it has information and then it's going to give us a custom tool where we're going to then spin up individual sessions s6 00:00:36.559 --> 00:00:38.320 for each of those financial artifacts. s7 00:00:38.320 --> 00:00:41.120 So you'll see this dispatch analysis thing. s8 00:00:41.120 --> 00:00:46.079 Um it's going to call it for two tickers and then it's going to kick runs for each of those. s9 00:00:46.079 --> 00:00:50.160 Um so this is how we can orchestrate our multiseession. s10 00:00:50.160 --> 00:01:00.480 um harness and each of those sessions has multi- aent uh that are doing collaborative work and each of those uh also have outcome definition so that there's self-supervision happening s11 00:01:00.480 --> 00:01:11.439 using memory stores to learn from one another um etc etc. Um so let's move on to the finale which is um what if you want to do recurring work? s12 00:01:11.439 --> 00:01:21.759 What if you want to do uh you want to have a list of companies you care about and you want every day at 9:00 am or first of the month to do the analysis over all of these companies s13 00:01:21.759 --> 00:01:22.799 uh on a schedule. s14 00:01:22.799 --> 00:01:24.000 So that's what deployments are. s15 00:01:24.000 --> 00:01:26.080 Deployments is everything is API based. s16 00:01:26.080 --> 00:01:29.840 So we you can obviously interact with our platform via the console. s17 00:01:29.840 --> 00:01:35.520 Um but their main way that we intend to you for you to interact with it is via the API and the SDK. s18 00:01:35.520 --> 00:01:39.040 So you can build uh custom UIs and custom workflows. s19 00:01:39.040 --> 00:01:45.280 Um all of these are just building blocks uh that you can compose together to build your application into something bigger. s20 00:01:45.280 --> 00:01:53.680 Um so you can see here what cloud did is it created a uh deployment um and that is right here. s21 00:01:53.680 --> 00:01:58.079 A deployment is takes in a agent definition. s22 00:01:58.079 --> 00:02:04.399 Uh in this case it's this research uh desk weekly memo using this specific agent. s23 00:02:04.399 --> 00:02:10.479 um it takes in a schedule uh of when it's going to run and it gives it an initial set of messages that it's going to run with. s24 00:02:10.479 --> 00:02:19.040 You can also pass in uh resources like uh uh memory stores or GitHub repos or files that it's going to use on each of those. s25 00:02:19.040 --> 00:02:25.280 Uh and so what this means is that every I think this is first of the month at 1 p.m. or something like that. s26 00:02:25.280 --> 00:02:30.160 It's going to run um this uh specific session. s27 00:02:30.160 --> 00:02:31.120 I could run one now. s28 00:02:31.120 --> 00:02:33.040 So I could trigger it via the console. s29 00:02:33.040 --> 00:02:34.480 it's going to do that. s30 00:02:34.480 --> 00:02:39.840 Um, but I could also build a UI for it and that's what we've done here. s31 00:02:39.840 --> 00:02:42.080 Um, so I can, you know, I have all the crowd. s32 00:02:42.080 --> 00:02:49.680 I can list and and search uh for my deployments and then I can also kick one off um at any given point. s33 00:02:49.680 --> 00:02:53.040 And I don't know why they're failing, but I'll ask Cloud about it later. s34 00:02:53.040 --> 00:02:56.959 Um, and this has come up a few times, which is what about dreaming? s35 00:02:56.959 --> 00:02:58.560 What about self-improvement? s36 00:02:58.560 --> 00:03:07.280 Um, so before I get to your questions, which I'm sure there are many, uh, I'm going to show you a preview that we've, uh, have out for for dreaming. s37 00:03:07.280 --> 00:03:15.599 So what dreaming is, it lets you select, um, so, okay, before I show you a previous one, it lets you select a memory store. s38 00:03:15.599 --> 00:03:19.120 In this case, it we'll we'll use the memory store that we've been using. s39 00:03:19.120 --> 00:03:20.800 You can select a model. s40 00:03:20.800 --> 00:03:25.280 Um, you can give it custom instructions if you want to, but we'll use a default prompt. s41 00:03:25.280 --> 00:03:32.400 Um, and then we're going to let it uh run uh the the last 30 days of sessions through this streaming analysis. s42 00:03:32.400 --> 00:03:47.519 So when I click start streaming, what it's going to do is it's gonna spin up a session using manage agents actually um spin up many sub agents to analyze the transcripts of each of the other sessions that we told it to evaluate at all of them which use the memory store. s43 00:03:47.519 --> 00:03:52.400 Um and it's going to analyze, hey, did the memory store encode all the information that was required? s44 00:03:52.400 --> 00:03:56.879 did a session uh run into any issue that uh that caused problems? s45 00:03:56.879 --> 00:03:58.400 Can we learn anything from it? s46 00:03:58.400 --> 00:03:59.840 Did this happen multiple times? s47 00:03:59.840 --> 00:04:02.080 Can we update the instructions in any way? s48 00:04:02.080 --> 00:04:17.440 You can see here what it actually did is that it found that the skill that we created actually had a few gaps um on the way that it was written and it updated the memory store to tell it even though there is this skill, you should extend this knowledge with the fact that we've learned from it. s49 00:04:17.440 --> 00:04:28.320 Uh and you can imagine that we will continue to extend this to not only improve memory stores but improve skills directly uh agent prompts, sub agent prompts, etc., etc. s50 00:04:28.320 --> 00:04:30.160 Um that's that's it. s51 00:04:30.160 --> 00:04:30.960 That's the demo. s52 00:04:30.960 --> 00:04:32.560 I sorry about the Wi-Fi. s53 00:04:32.560 --> 00:04:35.919 I hope you follow uh later on when when you get home. s54 00:04:35.919 --> 00:04:38.320 Um and let's uh keep going with some questions. s55 00:04:38.320 --> 00:04:40.960 Yeah, we can all take questions. s56 00:04:40.960 --> 00:04:41.840 We have a question over here. s57 00:04:41.840 --> 00:04:42.479 Yep. s58 00:04:42.479 --> 00:04:43.440 Go ahead. s59 00:04:43.440 --> 00:04:44.560 Uh thank you for the demo. s60 00:04:44.560 --> 00:04:49.280 Uh we didn't see any uh telemetry uh instrumentation in the agent. s61 00:04:49.280 --> 00:04:51.520 Is that a default native feature? s62 00:04:51.520 --> 00:04:52.800 What do you mean by telemetry? s63 00:04:52.800 --> 00:04:56.240 Because you have the full observability of all of the events uh that happened. s64 00:04:56.240 --> 00:04:56.479 Right? s65 00:04:56.479 --> 00:05:05.759 If you go into the console, you can see it live, but you can also obviously get it from the API where this transcript view is like a nice view, but if you go to debug, s66 00:05:05.759 --> 00:05:08.000 it's every event that occurred in that session. s67 00:05:08.000 --> 00:05:12.639 So you can see exactly, for example, when did we start calling the model? s68 00:05:12.639 --> 00:05:14.240 uh when did we finish calling the model? s69 00:05:14.240 --> 00:05:15.919 So you can get latency metrics in that. s70 00:05:15.919 --> 00:05:18.320 Is there something else that you mean by telemetry? s71 00:05:18.320 --> 00:05:18.560 Yeah. s72 00:05:18.560 --> 00:05:24.240 Uh the thing is that uh usually in software development cycle you kind of define what you want to observe. s73 00:05:24.240 --> 00:05:28.800 Uh how did we kind of pick what did we want to observe. s74 00:05:28.800 --> 00:05:32.960 Uh right now you get all the events and you just filter them on your end. s75 00:05:32.960 --> 00:05:37.759 If there are things that you think are missing, uh hit us up and we'll talk about adding them. s76 00:05:37.759 --> 00:05:38.000 Okay. s77 00:05:38.000 --> 00:05:41.680 And can I bring my own observability stack in here? s78 00:05:41.680 --> 00:05:43.919 Uh you mean like export to OTEL? s79 00:05:43.919 --> 00:05:44.240 Yeah. s80 00:05:44.240 --> 00:05:49.120 Uh you you could because it's an API you could build that connector fairly easily. s81 00:05:49.120 --> 00:05:56.400 Uh we just everybody has different requirements and we don't want to be opinionated but everything is just API calls. s82 00:05:56.400 --> 00:05:56.800 Mhm. s83 00:05:56.800 --> 00:05:58.000 Thank you. s84 00:05:58.000 --> 00:05:58.639 We have a question. s85 00:05:58.639 --> 00:06:00.639 Uh I I had a question. s86 00:06:00.639 --> 00:06:03.039 So how long can these agents run? s87 00:06:03.039 --> 00:06:15.440 I I I know you mentioned this dreaming thing but can you run them indefinitely or do you see that persistent store keep increasing in size and at some point you need to recompact it offline or something? s88 00:06:15.440 --> 00:06:19.680 We we are running compaction for you so that you can run these for days. s89 00:06:19.680 --> 00:06:23.759 We've seen 10-hour sessions in production uh from some of our users. s90 00:06:23.759 --> 00:06:27.919 So you could run these agents like for a year or indefinitely in theory. s91 00:06:27.919 --> 00:06:28.400 Yeah. s92 00:06:28.400 --> 00:06:33.360 Are you talking about running the agent for a year or you talking about like one specific session that keeps going for s93 00:06:33.360 --> 00:06:39.759 No, basically the agent if I had to if I had a task that I need to do every day, can I just keep it running for 6 months? s94 00:06:39.759 --> 00:06:42.400 Will it not blow its uh persistence? s95 00:06:42.400 --> 00:06:42.800 That's right. s96 00:06:42.800 --> 00:06:43.520 Yeah. s97 00:06:43.520 --> 00:06:54.800 And is there any orchestration piece like where I could have three agents, agent A does something and delegates to B and C. That's some capability that you have or s98 00:06:54.800 --> 00:06:58.960 for for now you would spin you would do sub agents and you would tell claude what you want. s99 00:06:58.960 --> 00:07:05.360 We find that uh there's a desire as humans to constrain what claude does and the way it does it. s100 00:07:05.360 --> 00:07:13.199 As cloud gets smarter we have continuously remove those guard rails and let claude just let it cook uh let it do what it thinks it's best. s101 00:07:13.199 --> 00:07:18.160 So uh right now you think maybe three sub aents is the best topology for it. s102 00:07:18.160 --> 00:07:23.039 But if we just let Claude figure that out, it actually ends up outperforming. s103 00:07:23.039 --> 00:07:30.000 But any topological or like graph based workflow thing in the UI here like where I could have s104 00:07:30.000 --> 00:07:30.479 Not yet. s105 00:07:30.479 --> 00:07:31.280 Not yet. s106 00:07:31.280 --> 00:07:31.759 Not yet. s107 00:07:31.759 --> 00:07:34.319 But but maybe soon. s108 00:07:34.319 --> 00:07:36.319 Uh s109 00:07:36.560 --> 00:07:38.639 you can probably just like speak up. s110 00:07:38.639 --> 00:07:39.039 Yeah. s111 00:07:39.039 --> 00:07:39.440 Sorry. s112 00:07:39.440 --> 00:07:40.160 Sorry. s113 00:07:40.160 --> 00:07:41.599 We'll come back to you. s114 00:07:41.599 --> 00:07:42.240 Sorry about that. s115 00:07:42.240 --> 00:07:45.199 Hey over here on your right. s116 00:07:45.199 --> 00:07:46.400 On my right. s117 00:07:46.400 --> 00:07:46.560 Okay. s118 00:07:46.560 --> 00:07:46.960 Yeah. s119 00:07:46.960 --> 00:07:47.680 So question. s120 00:07:47.680 --> 00:07:48.960 Um, looks great. s121 00:07:48.960 --> 00:07:50.560 A lot of integration points. s122 00:07:50.560 --> 00:08:03.199 What I'm curious about, uh, for, you know, an organ, a a complex organization that potentially manages multiple agents, um, you know, I I I think they want some sort of separation, s123 00:08:03.199 --> 00:08:04.800 arbback, whatever else. s124 00:08:04.800 --> 00:08:15.599 Um, how is that what is what is I guess the recommended approach for uh, a complex organization that has those different agents in here to effectively manage? s125 00:08:15.599 --> 00:08:27.120 We we have workspaces that's our primitive today where you can put different customers in different workspaces and then uh segment them that way and give access to specific users s126 00:08:27.120 --> 00:08:28.720 into those workspaces. s127 00:08:28.720 --> 00:08:36.159 Uh but we know that there's still a gap in what we offer and um that's that's something s128 00:08:36.159 --> 00:08:37.279 our back is on our road map. s129 00:08:37.279 --> 00:08:48.320 So we we do want to provide more like granular controls but right now workspace is like a clean way of like separating your agents where you have like a group of people access to workspace A and they can access all agents within that. s130 00:08:48.320 --> 00:08:54.000 You can have workspace B with like different set of users that can have different levels of access. s131 00:08:54.000 --> 00:08:57.040 So hi excellent presentation really excited. s132 00:08:57.040 --> 00:08:58.399 Uh two questions. s133 00:08:58.399 --> 00:08:59.440 One is deployment. s134 00:08:59.440 --> 00:09:06.640 My understanding was that a deployment is like at a snapshot a moment in time of all the agents. s135 00:09:06.640 --> 00:09:07.360 No s136 00:09:07.360 --> 00:09:12.160 deployment is essentially a way for you to run sessions for a specific agent on a schedule. s137 00:09:12.160 --> 00:09:22.399 So you can say like oh research desk agent go research NVIDIA once every quarter and it'll just create a session every quarter to go research and produce you that report. s138 00:09:22.399 --> 00:09:27.200 And then memory which is a state is persisted across all sessions. s139 00:09:27.200 --> 00:09:27.680 That's right. s140 00:09:27.680 --> 00:09:29.920 You attach your memory store to the session. s141 00:09:29.920 --> 00:09:34.320 So you decide which which session you want to attach which memory store too. s142 00:09:34.320 --> 00:09:38.720 So the memory store is like a independent entity that you just attach whenever you need to s143 00:09:38.720 --> 00:09:47.200 right and the final question is harnesses right one what is harnesses for you because everyone have their own definition s144 00:09:47.200 --> 00:09:49.519 and second is the confident right. s145 00:09:49.519 --> 00:09:54.800 So when I change something here, how do I know that it's not s146 00:09:54.800 --> 00:09:55.920 it's working as expected? s147 00:09:55.920 --> 00:09:56.160 Right? s148 00:09:56.160 --> 00:09:58.800 My own confident as a developer. s149 00:09:58.800 --> 00:10:02.720 Um for you're right, everybody has a different definition of a harness. s150 00:10:02.720 --> 00:10:09.120 Uh we when we talk about a harness in CMA, we mainly mean the tool loop execution. s151 00:10:09.120 --> 00:10:16.560 So managing the context and managing the session state and making sure that the agent can do the work that it wants to do. s152 00:10:16.560 --> 00:10:30.880 When most builders talk about harnesses, they they really mean, you know, they really mean um the ability to coordinate multiple agents and sessions together as a unit of work to achieve some sort of bigger task in a topology of sorts. s153 00:10:30.880 --> 00:10:41.279 Um we're giving you the abstraction of a single session as a unit of work for you to go out and build your harness uh with it as the session being one of the primitives. s154 00:10:41.279 --> 00:10:41.680 Um s155 00:10:41.680 --> 00:10:45.200 and then your second question about how do how do I know how my agent is performing? s156 00:10:45.200 --> 00:10:54.480 Our best recommendation would be to run evals on your agent with like different types of tasks that you want the agent to perform and see how the agent does on those tasks. s157 00:10:54.480 --> 00:11:10.320 Um outcomes is a really good way where we steer the agent for you to get you the specific outcome where you can define an exact rubric for like hey this is the exact outcome I want and you can define max iterations and the agent will keep going into it until it gets you that outcome. s158 00:11:10.320 --> 00:11:14.000 Yeah, the the the verifier agent that that Gabe talked about. s159 00:11:14.000 --> 00:11:16.880 Hi So s160 00:11:16.880 --> 00:11:17.920 after. s161 00:11:17.920 --> 00:11:18.480 Cool. s162 00:11:18.480 --> 00:11:19.600 Hey, thanks for the demo. s163 00:11:19.600 --> 00:11:20.720 I enjoyed it. s164 00:11:20.720 --> 00:11:30.640 Um my my question at the core is a bit about how to choose between skills and agents and I wanted to ask how would you all say what's like the fundamental difference? s165 00:11:30.640 --> 00:11:36.480 Is there an advantage to the harnessing that surrounds an agent that makes it produce better results than a scale? s166 00:11:36.480 --> 00:11:49.279 Um, and one of the reasons I'm asking is because I'm afraid of always investing in new kind of like I guess harnessing and tooling around models and it's it can be kind of tiring always trying to do a new thing. s167 00:11:49.279 --> 00:11:54.959 Um, so yeah, I just want to hear what you think is like is there a big difference and is it going to change any time? s168 00:11:54.959 --> 00:12:05.120 I yes, no. Um, the qu thick hardness, thin harness, thick skills um is is a question that's constantly evolving. s169 00:12:05.120 --> 00:12:14.560 I think what we find is that as Claude gets smarter, the boundaries and and and help that we want to give Claude start to melt away and disappear. s170 00:12:14.560 --> 00:12:18.000 And so what you need today might not be what you need tomorrow. s171 00:12:18.000 --> 00:12:22.639 The boring answer is just run evals and and evaluate these things. s172 00:12:22.639 --> 00:12:24.639 That sounds exhausting obviously like yeah s173 00:12:24.639 --> 00:12:27.920 every time a new model comes out I need to evaluate all these things. s174 00:12:27.920 --> 00:12:31.839 Um what what we find is skills is a good abstraction today. s175 00:12:31.839 --> 00:12:34.000 uh it gives the model enough s176 00:12:34.000 --> 00:12:35.839 information to do what it needs to do. s177 00:12:35.839 --> 00:12:43.600 But I presume that it will start to shift and melt and become thinner and thinner where the model will just be able to do things on its own without any s178 00:12:43.600 --> 00:12:47.200 special requests and instructions uh down the line. s179 00:12:47.200 --> 00:12:56.480 So uh multi-agent seems to be more useful uh in my opinion, but your your mileage may vary on your task. s180 00:12:57.040 --> 00:12:57.440 Hi. s181 00:12:57.440 --> 00:12:57.680 Hey. s182 00:12:57.680 --> 00:12:58.320 Hi. s183 00:12:58.320 --> 00:13:03.600 So my question ties up to last two questions and especially on top of evals. s184 00:13:03.600 --> 00:13:09.760 One thing I'm excited about is dreaming but how will dreaming actually affect the final outcome s185 00:13:09.760 --> 00:13:17.120 something that's where eval comes into the picture but you can have multiple iterations of dreaming sel that can go in different directions. s186 00:13:17.120 --> 00:13:30.160 So is there a plan or do you already support some automated kind of dreaming with run with evals and see that hey how to different direction go into which one should we pick based upon what's going on. s187 00:13:30.160 --> 00:13:30.399 Yeah. s188 00:13:30.399 --> 00:13:36.959 So so dreaming what it does is it looks over you you can choose the duration for which you want the agent to dream over. s189 00:13:36.959 --> 00:13:52.800 So you can say I want to dream over this agent session over the last 30 days and then look at all the transcripts and analyze what task you're trying to achieve and then come up with like a new memory store that's more geared towards like what you were trying to do and we typically will run evals s190 00:13:52.800 --> 00:14:07.279 on our end to make sure that dreaming is providing you with better outcomes and we've seen customers were able to get be get a significant improvement on their evals by running Dreaming and then using the new memory server but again everything kind of ties back to eval. s191 00:14:07.279 --> 00:14:10.480 So like that that's how you would know whether something is performing s192 00:14:10.480 --> 00:14:13.040 as you expect it. s193 00:14:20.240 --> 00:14:22.880 Yeah, I think it'd be good to have your own set of evals. s194 00:14:22.880 --> 00:14:23.839 Um s195 00:14:23.839 --> 00:14:26.399 the other thing is dreaming is in research preview. s196 00:14:26.399 --> 00:14:28.320 So these are early early. s197 00:14:28.320 --> 00:14:30.000 It's like a big experimental thing. s198 00:14:30.000 --> 00:14:32.480 We've seen uplift uh on it generally. s199 00:14:32.480 --> 00:14:36.720 This is we we hear you right like uh how do I gain the confidence that this is s200 00:14:36.720 --> 00:14:47.199 yeah it's like eval connected to dreaming and then we can find out multiple pathways that that has been taken and decide oh this makes more sense because even though s201 00:14:47.199 --> 00:14:50.800 nine sessions may have been saying that hey this is wrong s202 00:14:50.800 --> 00:14:59.120 one says that hey this is good so if you just rely on that session dreaming thing I feel like we may not be grabbing all insights without evalu s203 00:14:59.120 --> 00:15:01.290 if I tell you just to trust me is that no s204 00:15:01.290 --> 00:15:03.290 [laughter] s205 00:15:03.600 --> 00:15:05.839 Thank you. s206 00:15:06.000 --> 00:15:07.360 Yeah. s207 00:15:07.360 --> 00:15:07.680 Hi. s208 00:15:07.680 --> 00:15:08.800 Uh thank you very much. s209 00:15:08.800 --> 00:15:10.399 Um this is very impressive. s210 00:15:10.399 --> 00:15:20.639 Uh those of us who have been working with cloud code for a year or more um have been doing things that are similar to this but this was nice to see this in a really mature form. s211 00:15:20.639 --> 00:15:27.199 Um, one question I have is how much of this was just built on top of cloud code? s212 00:15:27.199 --> 00:15:33.519 That is if somebody was, you know, ambitious, could they could they build something similar to this? s213 00:15:33.519 --> 00:15:38.480 And um, what I wanted to ask you was uh, encourage you to keep things open. s214 00:15:38.480 --> 00:15:50.079 uh some of us are already harvesting what's in the JSON session files uh and starting to experiment with things like uh hepo and that kind of thing for for managing the um s215 00:15:50.079 --> 00:15:53.920 the memories and um and creating a memory pool. s216 00:15:53.920 --> 00:15:55.360 So uh congratulations. s217 00:15:55.360 --> 00:15:56.160 Thank you. s218 00:15:56.160 --> 00:16:04.800 Please keep it open and um can we is there are you doing anything here that's not just built on site on top of cloud code? s219 00:16:04.800 --> 00:16:06.000 Thank you. s220 00:16:06.000 --> 00:16:08.560 We we took the cloud code harness as an inspiration. s221 00:16:08.560 --> 00:16:12.000 We found that it was very well attuned for coding tasks. s222 00:16:12.000 --> 00:16:17.040 Um and different it's different levels of flexibility for other tasks. s223 00:16:17.040 --> 00:16:27.759 We're building an API where we want you to bring your context, your information, your capabilities and build on top of it where cloud code is trying to give you a a product out of the box. s224 00:16:27.759 --> 00:16:44.240 So we have slightly different uh you know value uh props um and so it's not just clocked out in a box uh though you know ideally it performs as well if not better and all sorts of tasks. s225 00:16:44.240 --> 00:16:51.920 Um so not not exactly but it's definitely inspired and motivated by by some of the uplifts that we have seen from cloud code s226 00:16:51.920 --> 00:17:01.920 and and cloud code is a terminal tool and we obviously want to be an API product that and infrastructure uh that is going to scale that's going to be reliable. s227 00:17:01.920 --> 00:17:10.480 So there there there were like challenges that I talked to you about um in the beginning of the presentation that we really worked in to make sure we're providing a really solid s228 00:17:10.480 --> 00:17:14.160 um product for you all to use. s229 00:17:14.400 --> 00:17:14.640 Yeah. s230 00:17:14.640 --> 00:17:20.640 Um, so for a bedrock, is there a console like clawed console like you have here? s231 00:17:20.640 --> 00:17:24.959 That's quite straightforward question, but yeah, I just want to manage and see everything. s232 00:17:24.959 --> 00:17:29.760 This is not available on bedrock, but it is available via cloud platform for AWS. s233 00:17:29.760 --> 00:17:36.240 So there is a AWS version of console that you could go use this with. s234 00:17:36.559 --> 00:17:37.760 Yep. s235 00:17:37.760 --> 00:17:38.240 Yeah. s236 00:17:38.240 --> 00:17:47.039 Uh first uh uh is there any uh how you say is there any restriction on the model to be used like uh non-anthropic models s237 00:17:47.039 --> 00:17:47.280 uh s238 00:17:47.280 --> 00:17:51.200 just like cloud code I can run on nonanthropic models s239 00:17:51.200 --> 00:17:52.559 for for now we we you s240 00:17:52.559 --> 00:17:54.799 yeah it's cloud models only for now s241 00:17:54.799 --> 00:18:03.200 okay so so the second question is what about end user use uh do they have to have a console accounts in order to use it or just a free s242 00:18:03.200 --> 00:18:07.039 uh I can create a customized UI for this and they can use it s243 00:18:07.039 --> 00:18:11.039 so cloud Cloud Console helps you with creating your API key. s244 00:18:11.039 --> 00:18:13.360 So that's how you get access to our APIs. s245 00:18:13.360 --> 00:18:16.080 So you cloud you're not paying separately for cloud console. s246 00:18:16.080 --> 00:18:17.200 It's free to sign up. s247 00:18:17.200 --> 00:18:20.799 You create an API key and then you're paying based on like token usage. s248 00:18:20.799 --> 00:18:27.039 So for for for the case like uh SAS like uh maybe we have multiple tenants right. s249 00:18:27.039 --> 00:18:42.799 So in order to provide this uh agent what kind of a uh how you say model you have already uh depict on so that we can follow for example what kind of a each company are supposed tenant are supposed to have a uh a different space or s250 00:18:42.799 --> 00:18:48.160 or even there is a a sub module for this uh each tenant. s251 00:18:48.160 --> 00:18:51.600 So are are we going to how do we actually going to organize that? s252 00:18:51.600 --> 00:18:52.559 In terms of this s253 00:18:52.559 --> 00:19:03.360 we we we have seen that the best model for multi-tenency is not bring your own key because you lose observability because it runs on your end users API key rather than than yours. s254 00:19:03.360 --> 00:19:08.240 So we encourage you to model things with workspaces as a tenency primitive. s255 00:19:08.240 --> 00:19:15.600 Um and we're improving that so that it's easier and easier to build uh on top of this for multi-tenant applications like s256 00:19:15.600 --> 00:19:26.240 yeah so I think currently you could either have put all your customers data into like a single workspace and then you maintain a mapping on your end on like which agent maps to which of your customers. s257 00:19:26.240 --> 00:19:34.400 Alternatively like Gabe said you could create a workspace per customer of yours but that can get more complicated because then you have to like recreate your agent. s258 00:19:34.400 --> 00:19:45.360 So, so is any you see the examples that uh people did the additional security layer or RBAC layer over the this workspace because to me it may be a little bit you know s259 00:19:45.360 --> 00:19:52.480 yeah that that would probably be like a good good way to do it right now because we don't have any arvback s260 00:19:52.480 --> 00:20:09.039 uh controls uh on our level other than like workspace because everything within your workspace is accessible to everyone in the workspace but we are working on making our back better and then making the experience better if you're working with cloud manage agents but building on behalf of your customers and how do we make that easy something s261 00:20:09.039 --> 00:20:16.640 so so is there any way you can show how to scale this uh if there's a multiple uh how you say tenant working on that s262 00:20:16.640 --> 00:20:17.039 yes s263 00:20:17.039 --> 00:20:26.400 any scaling parameters like uh what is the how you say the the minimal uh pod or whatever these parameters that's was yeah s264 00:20:26.400 --> 00:20:40.000 uh I mean you're you're defining on a you can create as many sessions as you want yeah like there's no limitation I mean we have some rate limits obviously but um but there's no scaling parameter for you to say now I need to run 10,000 pods s265 00:20:40.000 --> 00:20:50.000 so so currently no no such thing as like uh like HPA in the Kubernetes like you can do these kind of setting scaling uh parameters right like timeout or whatever s266 00:20:50.000 --> 00:20:51.520 you you don't need that s267 00:20:51.520 --> 00:20:55.520 you don't need to because we're managing it all for you you're just making API calls that's all s268 00:20:55.520 --> 00:21:05.039 okay okay sure uh thanks for your uh answer the last uh by the way I the offering link does not work for me I'm not sure if it's the only one. s269 00:21:05.039 --> 00:21:05.440 Yeah. s270 00:21:05.440 --> 00:21:06.320 Thank you for the sharing. s271 00:21:06.320 --> 00:21:06.799 Thank you. s272 00:21:06.799 --> 00:21:08.559 Thank you. s273 00:21:08.559 --> 00:21:08.880 Yeah. s274 00:21:08.880 --> 00:21:09.679 Go ahead. s275 00:21:09.679 --> 00:21:10.720 Hi. s276 00:21:10.720 --> 00:21:11.840 First, thank you very much. s277 00:21:11.840 --> 00:21:18.240 I've been having a ball with everything that you've been releasing and um I'm really interested in how you're using UI. s278 00:21:18.240 --> 00:21:29.360 Um you're spinning up a server, you're distributing this, but how might we deploy that either internally, but I'm interested in to sort of a customer base. s279 00:21:29.360 --> 00:21:33.600 If I want to make for example an astrologer s280 00:21:33.600 --> 00:21:36.559 um that person has to enter sensitive data. s281 00:21:36.559 --> 00:21:39.600 So it's actually really nice for me to be able to like have that in a s282 00:21:39.600 --> 00:21:40.240 have a layer. s283 00:21:40.240 --> 00:21:40.799 Yeah. s284 00:21:40.799 --> 00:21:44.559 So there are some procedural things you get like a star chart example. s285 00:21:44.559 --> 00:21:52.880 I'm not building this but it's a good example like I can have a sub agent do an interpretation that could last for like 90 seconds. s286 00:21:52.880 --> 00:21:57.280 It could create some assets etc. When the outcome is met it could shut down. s287 00:21:57.280 --> 00:22:03.440 So this gives me a way to actually just make a really clean implementation where all I need is off s288 00:22:03.440 --> 00:22:05.600 and then everything can spin up from there. s289 00:22:05.600 --> 00:22:06.159 Yeah, exactly. s290 00:22:06.159 --> 00:22:16.080 So there are a couple of things like MCP channels that kind of point to this with like fake chat for example, a way for you to kind of just use claude s291 00:22:16.080 --> 00:22:19.919 and have a channel to communicate with it through UI. s292 00:22:19.919 --> 00:22:25.760 But I'm seeing a lot where you're just spinning up a server inside a project that's easy to distribute internally. s293 00:22:25.760 --> 00:22:28.880 But how can you deploy that at scale? s294 00:22:28.880 --> 00:22:33.360 Um I mean we've this is uh being deployed today at scale in production. s295 00:22:33.360 --> 00:22:36.480 For example, notion external agents are built on top of this. s296 00:22:36.480 --> 00:22:47.200 So this is something that exists today that you can go and use and uh basically leverage cloud managed agents through notion in Atlassian for Jira integrations for example. s297 00:22:47.200 --> 00:22:48.960 Uh there's definitely ways to do it. s298 00:22:48.960 --> 00:22:51.840 I think if you have specific questions we could talk uh s299 00:22:51.840 --> 00:23:01.039 yeah we we can talk after this but typically you would probably build your product your UI that you're distributing to your customers and then calling into us for like your actual agent s300 00:23:01.039 --> 00:23:02.159 the phrase s301 00:23:02.159 --> 00:23:04.159 management everything is through the API s302 00:23:04.159 --> 00:23:06.000 everything that's right s303 00:23:06.000 --> 00:23:12.320 so you will manage all the API calls and build your product and UI around that yeah s304 00:23:12.400 --> 00:23:13.520 hi s305 00:23:13.520 --> 00:23:20.799 hi hi first thanks for a great workshop and demo and uh my question is Not really to like the agent itself. s306 00:23:20.799 --> 00:23:24.480 It's just like I saw the Android just released the sunet five. s307 00:23:24.480 --> 00:23:24.799 Yes. s308 00:23:24.799 --> 00:23:30.080 So I just was wondering if any anything you guys could share with us. s309 00:23:30.080 --> 00:23:31.200 Yeah. s310 00:23:31.200 --> 00:23:32.159 It's a great model. s311 00:23:32.159 --> 00:23:32.851 Yeah. s312 00:23:32.851 --> 00:23:34.159 [laughter] s313 00:23:34.159 --> 00:23:36.400 Try it. s314 00:23:37.760 --> 00:23:40.320 It is supported in cloud manage agents. s315 00:23:40.320 --> 00:23:41.360 Go give it a try. s316 00:23:41.360 --> 00:23:43.039 You can use it today and see. s317 00:23:43.039 --> 00:23:43.600 Hey there. s318 00:23:43.600 --> 00:23:53.840 Um hi if we've got uh business partners that are using like cloud co-work today some of them more techsavvy cloud code to build some like spot solutions for specific business cases. s319 00:23:53.840 --> 00:24:02.400 What do you see as sort of the level of sophistication up level to start using man like managed agents instead because this seems like more powerful s320 00:24:02.400 --> 00:24:06.960 but the you know cloud platform could be intimidating to folks that aren't engineers let's say. s321 00:24:06.960 --> 00:24:17.600 Yeah, I think the console quick start is probably like the part of my demo where you're literally just like telling Claude the agent that you're trying to build should make that pretty easy for s322 00:24:17.600 --> 00:24:20.559 even non-technical users to come in and build their first agent. s323 00:24:20.559 --> 00:24:25.039 But obviously this is an API product and is naturally going to be a little bit more intimidating. s324 00:24:25.039 --> 00:24:33.919 But if they're comfortable with cloud code, we do have a manage agent skill with cloud code that basically just like in natural language will walk you through building your first agent. s325 00:24:33.919 --> 00:24:37.679 And that would probably be my advice uh to try it out. s326 00:24:37.679 --> 00:24:40.960 I think the biggest difference is co-work building building personal agent. s327 00:24:40.960 --> 00:24:49.200 With this one, you're trying to build probably an agent that's like available to like your team, your internal organization, or you're building a product that you're shipping to your end customers. s328 00:24:49.200 --> 00:24:54.240 But if you're building personal agents, co-work is still a really great tool to continue doing that. s329 00:24:54.240 --> 00:24:58.799 Are there any skills to kind of like take something that exists in co-work? s330 00:25:01.279 --> 00:25:04.320 there's a skill to interact with our API. s331 00:25:04.320 --> 00:25:06.480 So that cloud is probably good enough. s332 00:25:06.480 --> 00:25:14.320 So if you say here's what I did in co and here's the skill to call the API make the translation I feel pretty confident. s333 00:25:14.320 --> 00:25:17.840 Yeah five we'll we'll do it for you. s334 00:25:17.840 --> 00:25:20.640 Um my question is about enterprise. s335 00:25:20.640 --> 00:25:33.760 uh there are c lot of enterprises who'd like to run these things inhouse because they have sensitive data sets that they won't want going out also they may have already set up their own secure proxies to anthropics models s336 00:25:33.760 --> 00:25:45.520 um GIS and also the other question was about scaling so within the enterprise now you can actually allocate compute to scale it our own way right uh so an on-prem kind of solution s337 00:25:45.520 --> 00:25:47.039 have you all thought about it s338 00:25:47.039 --> 00:26:01.360 uh definitely been a customer ask from a bunch of for enterprise customers and and we're seriously thinking considering it at the moment, but we we don't already have something that we can we can share with you, but we're we're we're seriously considering it. s339 00:26:01.679 --> 00:26:02.240 Excuse me. s340 00:26:02.240 --> 00:26:08.480 Could you show the last slide that show the QR code of the workshop? s341 00:26:08.480 --> 00:26:11.760 Uh are you looking for the repo or the credits? s342 00:26:11.760 --> 00:26:12.400 The repo. s343 00:26:12.400 --> 00:26:12.720 Okay. s344 00:26:12.720 --> 00:26:14.720 Yeah.